I’m using VXL through ITK and am looking at speeding up matrix inversion (for big matrices ~5000*5000). I’m using VXL 1.9.0 (from ITK 3.20). I hope my question is not too outdated since the current version is 1.13.
From what I understand, the SVD implemented in VNL uses LINPACK. Isn’t LAPACK faster? Since LAPACK is already included in VXL, are you planning on using it for SVD? Is there (or will there be) a way to switch from one implementation using LINPACK to another using LAPACK if people still want to use LINPACK?